This paper from the Kellis lab is a great demostration of how functional genomic studies should be in the post hight throughput sequencing era.
They took on a classic GWAS example, obesity, which has a fairly high genetic component, up to 80% heritability. A 47kb region spanning the first and second intron of the FTO gene is strongly associated with the genetic variation of body-mass index (BMI). Subsequent KO studies in animal models revealed a direct functional link between the FTO protein and control of thermogenesis, energy expenditure and regulation of body weight.
Then you wanna ask yourself: WTH are we still doing, when FTO sure looks like the culprit for obesity? Because none of the associated variants could explain the disregulation of FTO expression. In other words, there’s a missing link between the genotype and phenotype, which puts the whole idea of FTO being causal into question.

Well, it turns out that, from an earlier work and confirmed in this study, a major gene product regulated by these disease variants are two genes, IRX3 and IRX5, located megabases away from the FTO locus. Both studies described in detail and convincingly about this long-range enhancer mechanism at the DNA and RNA level.
Extended from this observation, this paper dissected out that IRX3 and IRX5, both transcription factors, regulate energy balance via their actions in adipocytes. They went on and showed that ARID5B is a master upstream regulator for both genes and concluded that one variant in the FTO locus, which serves as an ARID5B binding site, is a causal genetic mutation of BMI, via the ARID5B-IRX3/5 module.
So, what’s up with all the FTO stuff? Is this gene still relevant in this context? Is it genetically connected with the IRXs? I asked the same question when commenting on that earlier paper and I couldn’t come up with a good explanantion. I think I know now. The strategy of making the KO mouse was to replace the majority of the endogenous FTO gene with a Neo cassette, leaving only exon1 and a fraction of intron 1 which doesn’t produce a functional protein. It is true that there’s no FTO expression in these KO animals, but a big chunk of the regulatory element in intron 1 and 2, as being described in both papers, is also eliminated. Therefore, all of the KO phenotypes could very well be due to the missing enhancer and misregulation of the IRX genes.
Now that is a big problem as almost all the KO mice are generated in this way. When you think you’re studying your favorite gene, think about all the introns and alternative exons you knocked out at the same time. All the enhancers, repressors, insulators, lincRNAs from the sense strand and the antisense strand, gone. Pretty scary, huh.